Statistics are collected based on page title rather than based to page ID (see T159046). This is a design bug yet to be resolved. An undesired result is inconsistent Massviews statistics when pages in categories or other collections are moved. Only the new title is searched in the historic data, where the old title, no longer in the collection, is discarded. This problem causes retroactive decrease in the collected page-view statistics.
- Mentioned In
- T121912: Better redirect handling for pageview API
- Mentioned Here
- T157830: Implement a Node.js backend for each Pageviews app
T141332: Query page move log to get pageviews of an article at its older locations
T159046: Track page views by page ID rather than title (handles moved pages)
T163621: Add option to include redirect views in pageviews
Despite what I wrote at T141332, it's actually pretty complicated. Instead we should just include all redirects, as the title of this task suggests.
The issue is Massviews can query tens of thousands of pages, and each can any number of redirects, increasing the total query count by an order of magnitude. That amounts to an enormous amount of data that may exhaust the client's memory. We might just include a disclaimer, or look into doing T157830 first so that memory consumption is done server-side.
In theory, redirects may increase the total number of pages by an order of magnitude. However, in practice, the average number of redirections per page is relatively small. For example, in my project a lot of pages have multiple redirections (alternative spelling and such). The total number of pages in the project category is 1190. The total number of pages including redirections is about 2000.